Why we're building Superposition

Three things I believe

I started Superposition because of three things I believe about engineering and AI.

First, smaller engineering teams can take on bigger ideas, and move faster on them, by using AI.

Second, engineers can only hand real work to AI if every decision it makes can be audited. An engineer signs their name to a design. They need to see where each number came from, and so does whoever reviews it after them.

Third, people and AI need one system to work in together, and it has to move as fast as the ideas do. That's what I mean by an operating system. Requirements, architecture, schedule, cost, risk and change sit in one plan, and people and agents work on it together.

Where we're going

Agents will do everything engineers need them to do. Engineers take on the rest: the hard problems, the judgment calls and the sign-off. All of it will stay auditable and traceable, so anyone can see who made each change and why.

We're not there yet. Today, an agent reads the plan and suggests changes, and a person decides each one. The list at the end says what works now and what comes next.

How it stays auditable

Connect Claude, or any AI assistant that supports MCP, the open standard for connecting AI tools. Each change it suggests waits in the Inbox with its reason, until a person accepts or rejects it. The record shows who suggested it and who decided.

Nothing is overwritten. A baseline changes only through a reviewed change, and the old one is kept.

Scheduling, costing, modeling and approvals never call an AI model.

Where it stands today

The capabilities page labels everything else, and the roadmap says what comes next.